Prognostic and predictive value of EGFR and EGFR-ligands in blood of breast cancer patients: a systematic review

Eligibility criteria

Studies assessing plasma or serum levels of EGFR or any EGFR-ligand at any time during breast cancer diagnosis, treatment or follow-up were included. Studies of breast cancer patients at any age, with any subtype and any stage of disease, as well as studies including patients receiving any breast cancer treatment, were included. Only studies investigating protein expression of EGFR or EGFR-ligands were included, whereas studies regarding gene expression were excluded. Studies reporting any kind of predictive or prognostic outcome measures were included, as well as all study designs and all durations of follow-up. Only data reported in peer reviewed journals were included and there were no restrictions to publication language or year of publication.

Sources and search strategy

Studies were identified by searching PubMed (1946–2016.08.29), Cochrane Library (1994–2016.08.29) and EMBASE (1974–2016.08.29). The search strategies were developed in a cooperation of three reviewers (TB, JSM, IMK) and a Health Sciences Librarian with expertise in systematic review searches. Both index terms (e.g. MESH terms) and text words were included in the searches. To ensure high sensitivity in relation to identification of relevant studies, the final search-string was broad and thus a priori expected to have relatively low specificity. A draft of all searches can be found in the Supplemental Material 1. EndNote X7™ was used to identify and remove duplicates and book sections. To ensure literature saturation, relevant publications were identified in the reference lists of the included studies and relevant reviews (IMK), and included in the study selection process.

Study selection

The initial selection of studies was performed blinded and solely based on assessment of the title of the publication. It was performed independently by two reviewers (TB, IMK). If one or both reviewers assessed the titles as relevant for abstract screening, they were included. The same two reviewers (TB, IMK) then independently assessed the abstracts and publications were included for full text assessment if one or both reviewers assessed them to be potentially relevant. Two reviewers (TB, IMK) independently assessed full text reports and included studies meeting the eligibility criteria. Any disagreement or doubt was resolved by achieving consensus through discussion among three reviewers (TB, JSM, IMK). The selection process is illustrated in Figure 1.

Figure 1:

Flow diagram illustrating the identification, screening and eligibility assessment of studies in accordance to Preferred Reporting Items for Systematic Review and Meta-analysis.

Data extraction

Data were extracted from each of the included studies to a data extraction form (Supplemental Material 2) by one reviewer (IMK) and subsequently verified by two other reviewers (TB, JSM). Data regarding study design, population, size, country, year and follow-up were extracted as well as patient characteristics (age, gender, receptor status, TNM-stage, grade, BRCA-mutations and treatment). Type of blood specimen and the investigated EGFR-ligands and/or EGFR were registered as well as the outcome measures and the results of the studies. Discrepancies were solved by thorough examination of the publications and subsequent consensus (TB, JSM, IMK).

Bias assessment

The risk of bias was assessed by two reviewers (TB, IMK) for all included studies, using a tool for assessing bias in studies of prognostic factors (“QUIPS tool”) [29]. Thus, for all included publications the risk of bias was assessed in six categories and the overall risk of potential bias was assessed. The six categories included study participation, study data available, prognostic factor measurement, outcome measurement, study confounding and statistical analysis and reporting. The risk of bias in each category was assessed as either high, moderate or low.

Data synthesis

Data synthesis was performed in collaboration with a statistician with expertise in meta-analysis in order to determine if the data were suitable for meta-analysis or if a narrative presentation would be preferred.

Literature search and study selection

The search identified 864 publications in PubMed, 1765 in Embase and 98 in Cochrane Library, a total of 2727 publications. After removal of duplicates (n=513) and book sections (n=62), 2152 publications were included for title screening. In the title screening 1960 publications were excluded leaving 192 publications for initial abstract screening. Additional 15 studies were identified through the reference lists of included papers and relevant reviews, and included in the abstract screening. In total 207 abstracts were screened and 168 publications were excluded after abstract screening. A total of 39 full text publications were assessed for eligibility and 23 were excluded due to the reasons outlined in Figure 1. Sixteen publications were included in the systematic review, of which only one was identified in the reference lists. Characteristics of the 16 included publications are shown in Table 1. A total of 12 studies evaluated EGFR and the results from these are shown in Table 2. Regarding the EGFR-ligands, AREG was evaluated in two studies, EGF in four studies and TGFα in two studies. Results are shown in Table 3. No studies investigating BTC, HB-EGF, EREG, EPGN or NRG-1-4 were identified. The biomarkers were analyzed in serum in all studies.

EGFR

S-EGFR was evaluated in relation to prognostic or predictive outcome measures in a total of 12 studies. Results are shown in Table 2. Of the 12 studies evaluating the predictive or prognostic value of S-EGFR, nine studies included patients with locally advanced or metastatic breast cancer [9], [20], [32], [33], [35], [36], [38], [40], [42], whereas two studies included patients with early breast cancer [37], [39]. In one study, patients with all clinical stages of disease were included and results were reported for the entire group of patients. Thus, in this study no specified information was provided regarding patients with metastatic breast cancer or early breast cancer [41].

EGFR-ligands

All five studies identified that examined the prognostic or predictive value of EGFR-ligands included patients with locally advanced or metastatic breast cancer, whereas no studies investigating the prognostic or predictive role of EGFR-ligands in early breast cancer were identified. Results are shown in Table 3. Furthermore, all studies that evaluated EGFR-ligands included patients with HER2-positive tumors only, except in one study in which HER2-status was unknown [31]. HR-status of the patients in the identified studies were either mixed or not reported as shown in Table 3.

Two of the five studies evaluating EGFR-ligands evaluated the value of S-AREG [26], [30]. One study found that patients with high S-AREG had significantly shorter PFS than patients with low S-AREG, whereas the other study found no correlation between S-AREG and either PFS or disease progression [26], [30].

A total of four studies evaluated the value of S-EGF [30], [31], [34], [36]. No correlations between S-EGF and either disease progression or PFS were found in one study [30], whereas another study reported that patients with high S-EGF had significantly longer OS compared to patients with low S-EGF, but no correlation to PFS was found [34]. One study reported that S-EGF did not identify patients with better or worse 2-year survival [31] and one study reported that there was no association between S-EGF and response [36].

Two studies evaluated S-TGFα [30], [36]. One study found the RR in patients with low S-TGFα to be significantly higher than in patients with high S-TGFα, whereas no correlations between S-TGFα and either TTP or OS were found [36]. The other study found no correlations between S-TGFα and either disease progression or PFS [30].

S-EGFR/EGFR-ligands in relation to S-HER2

Several of the identified studies evaluated both S-HER2 and S-EGFR as prognostic or predictive biomarkers in breast cancer; however, only three of the identified studies reported a combined analysis of these two biomarkers [20], [40], [42]. Souder et al. found no added value from S-EGFR when compared to S-HER2 alone in relation to ORR, CBR, TTF and TTP. However, patients with decreased S-EGFR (<44.1 ng/mL) and normal S-HER2 (<15 ng/mL) had significantly reduced OS compared to patients with normal levels [40], [43]. Witzel et al. [42] and Muller et al. [20] evaluated the combination of S-EGFR and S-HER2 in metastatic breast cancer as well, and both studies reported a tendency of shorter survival in patients with high S-HER2 and low S-EGFR as compared to patients with normal levels of both markers. Regarding serum levels of EGFR-ligands, no studies evaluated the combination with S-HER2.

Bias assessment

Assessment of risk of bias can be found in Supplemental Material 3. Overall, the risk of bias in the included studies varied. A common and crucial risk of bias was lack in data reporting and inadequate description of study populations. The risk of confounding was not accounted for in several of the studies, i.e. HER2-status of the tumor. Outcome measures were well defined in most studies. Finally, several studies did not report either the coefficient of variation or the limit of detection of the biomarker analysis (Supplemental Material 3).

Data synthesis

The 16 studies included in this systematic review showed considerable clinical heterogeneity both in relation to differences in study designs, outcome measures and in relation to differences in characteristics of study subjects, such as stage and subtype of breast cancer. In addition, the studies were heterogeneous when it comes to interventions both in relation to type of treatments and dosage and durations of treatments. Thus, the synthesis of the current evidence is narrative [44].

EGFR

The 12 studies evaluating the prognostic or predictive value of S-EGFR showed considerable variation in relation to study design, patient characteristics (disease stage, HER2-status, HR-status) and treatment regimens. In addition different cut-off levels for S-EGFR were used and the reasons for choice of cut-off varied between studies or were not explained. However, all significant correlations observed in advanced and metastatic breast cancer showed an overall association between low baseline S-EGFR and a shorter survival or reduced response to treatment [20], [38], [40]. Though several studies did not find the same significant trends, none of the identified studies reported opposing results. To our knowledge, no clear explanations for the observed correlations are known. It has, however, been hypothesized that as part of maintaining the proliferative activity, cancer cells with increased malignant potential may have a decreased proteolytic cleavage of the extracellular part of EGFR and studies have reported that levels of S-EGFR are lower in metastatic breast cancer patients, as compared to healthy controls [20], [21]. The associations between low baseline S-EGFR and shorter survival or reduced response to treatment in metastatic breast cancer might also be a reflection of this. However, it is unknown if the correlations observed reflect prognostic information regarding malignity or predictive information regarding response to treatment. In the studies by Souder et al. [40] and Muller et al. [20], the correlations between low S-EGFR and reduced survival were found only in patients with HR-positive tumors/HR-status unknown who received anti-hormonal treatment and cytostatic treatment, respectively. This might indicate a possible prognostic or predictive role of S-EGFR, especially in HR-positive breast cancer patients. It is, however, unknown if the correlations might reflect associations between EGFR-expression and ER-expression in relation to prognosis in recurrent breast cancer [18], [45] or associations between EGFR-expression and response to treatment.

Overall, current evidence indicates that low S-EGFR in metastatic breast cancer is associated with shorter survival and reduced response to treatment, especially in HR-positive patients. However, findings are not consistent across studies and comprehensive aspects of the utility of S-EGFR as a predictive and prognostic marker in breast cancer remain to be uncovered.

EGFR-ligands

The prognostic and predictive value of S-AREG, S-EGF and S-TGFα in breast cancer has been investigated in several studies, whereas no studies evaluating the value of S-BTC, S-HB-EGF, S-NRG, S-EREG or S-EPGN were identified (Table 3). All studies regarding EGFR-ligands were performed in patients experiencing advanced or metastatic breast cancer and all the included patients were HER2-positive, except in one study where HER2-status was not reported. Baselga et al. did not find S-AREG, S-EGF or S-TGFα to be associated with PFS in any of the treatment groups. However, despite levels of the ligands in serum are known at baseline, week 9 and at progression, the possible information of a decrease or an increase in tumor marker levels were not evaluated [30]. Evaluation by cut-offs might limit the sensitivity of the tumor marker as compared to evaluation by changes in levels of the tumor markers from baseline [46]. Several other studies evaluated the value of S-AREG, S-EGF and S-TGFα as well [26], [31], [34], [36]. These studies, however, included only few patients and the results should be interpreted with caution. Overall, S-AREG, S-EGF and S-TGFα have only been partially investigated in highly selected subsets of breast cancer patients, whereas the prognostic and predictive value of the remaining EGFR-ligands has not yet been explored in a clinical setting. Evidence towards the importance of EGFR-ligands as biomarkers in breast cancer is provided by studies reporting aberrant expression of some EGFR-ligands in breast cancer tissue and associations between expression of EGFR-ligands in tissue and prognosis [16], [47]. Furthermore, several case-control studies show higher levels of EGFR-ligands in the blood of breast cancer patients as compared to healthy controls [22], [23], [24], [25]. Thus, further studies are needed in order to clarify the role of EGFR-ligands as prognostic or predictive markers in breast cancer patients as well as for monitoring effect of treatment and/or resistance.

S-EGFR/EGFR-ligands in relation to S-HER2

A crosstalk between HER2 and EGFR is known to be crucial in the receptor signaling, and HER2 is the preferred heterodimerization partner of EGFR [4]. Furthermore, there are indications that EGFR and EGFR-ligands are involved in resistance to trastuzumab [4]. S-HER2 is known to provide valuable information regarding response to trastuzumab in HER2-positive breast cancer [13], [48]. Though S-HER2 is not the focus of this review, it is relevant to evaluate, if combined analysis of serum levels of HER2 and EGFR or EGFR-ligands provide additional clinical relevant information. Only three studies reported a combined analysis of both S-EGFR and S-HER2 in patients with metastatic breast cancer [20], [40], [42], whereas no studies reported a combined analysis of EGFR-ligands and S-HER2. Two of the studies [20], [42] that evaluated the combination of S-EGFR and S-HER2 reported a tendency of shorter survival in patients with high S-HER2 and low S-EGFR as compared to patients with normal levels of both markers. However, it is unclear whether the results of these two studies reflect the well examined effect of S-HER2 and/or S-EGFR since both studies reported only results of patients with normal levels of both markers as compared to patients with aberrant levels of both markers [20], [42]. Another study, however, reported that patients with decreased S-EGFR (<44.1 ng/mL) and normal S-HER2 (<15 ng/mL) had significantly reduced OS compared to patients with normal levels [40], [43]. Overall, there are indications that low S-EGFR may be associated with shortened survival in patients with normal levels of S-HER2. This might be partially explained by a resistance mechanism due to a crosstalk between EGFR and ER in HR-positive patients receiving endocrine therapy. Further investigation of the combination of serum levels of HER2 and EGFR/EGFR-ligands is warranted.

Perspectives

In recent years, HER2-targeted treatments have radically improved the prognosis for patients with HER2-positive tumors. As dual and pan-targeted therapies against the EGFR-system are now continuously implemented in breast cancer treatment, it is highly desirable to identify markers or a combination of markers that might predict response or resistance to the treatments. In lung cancer and colorectal cancer, EGFR-targeted treatments have an established therapeutic role [49], [50] and there are indications, that levels of EGFR and some of the EGFR-ligands in blood can provide valuable information regarding prediction of treatment outcome or prognosis in certain groups of patients [51], [52], [53], [54], [55]. In breast cancer, important aspects of the clinical implications of EGFR and EGFR-ligands in blood remain to be explored.